Skip to main content
NOPE
Medium reported severity Regulatory Action

New Mexico Supreme Court fines attorney Stephen Aarons $5,000 and removes him from a murder appeal over ChatGPT-fabricated testimony

Santa Fe defense attorney Stephen Aarons used ChatGPT (OpenAI's o3) on a Rev.com transcript and the record to draft briefing in the murder appeal of Oscar Renee Sandoval; the brief contained fabricated witnesses and testimony (including police officers 'Michelle Amarillo' and 'Sanchez' and invented testimony attributed to real people) and misdescribed real precedents. After a 21 August 2026 show-cause hearing at which Aarons admitted he had not verified the output, the New Mexico Supreme Court's written order of 9 September 2026 held him in direct contempt, fined him $5,000 payable to the State Bar Client Protection Fund within 30 days, removed him from the case, barred him from appearing before the court pending a disciplinary-board referral, struck the filed briefs and reassigned the appeal to the public defender.

AI system
ChatGPT
OpenAI
Occurred
1 Aug 2025 to 9 Sept 2026
Reported
24 August 2026
Event location
United States
What the AI did
Communicated with the person
Reported harm
Professional HarmLegal HarmFinancial Loss
Whose AI use
Their own AI use
Setting
Work · Justice
Evidence
AI involvement supported · Causal attribution established · 3 sources
5 claims: 2 corroborated, 3 reported. 5 open questions
People reported harmed
2 people

AI system as recorded: ChatGPT (OpenAI o3)

What Happened

Aarons filed the brief around August 2025 in the appeal of Oscar Renee Sandoval, sentenced to life in February 2025 in Doña Ana County for killing his partner Shiereen Al-Jibury. The State moved to strike weeks later. Reporting by Ars Technica (11 September 2026), quoting the order and the hearing, records that Aarons fed a Rev.com transcript and the case record into ChatGPT (OpenAI o3) and that the resulting brief contained fabricated witnesses and testimony — officers 'Michelle Amarillo' and 'Sanchez', Manal Al-Jibury and Teresa Marquez — and false testimony attributed to Danny and Linda Stanton and Mariah Chavez, alongside inaccurate descriptions of real precedents. At the 21 August 2026 show-cause hearing he admitted the errors and his failure to verify (the ABA Journal reported the hearing and the anticipated $5,000 contempt on 24 August, citing the Santa Fe New Mexican). The 9 September 2026 written order held him in direct contempt, imposed the $5,000 fine payable to the Client Protection Fund within 30 days, removed him from the case, barred him from appearing before the court pending the disciplinary-board referral, struck the briefs and assigned the appeal to the public defender (Kim Chavez Cook, appointed 2 September per Ars). The court told Aarons 'your client is the one suffering': Sandoval's appeal is delayed while the substitution proceeds. The order PDF itself could not be retrieved in this review (nmcourts.gov search returned nothing; the case lookup is form-based); the order's contents are quoted extensively by Ars Technica and confirmed by the Reuters wire.

Reported harm

A dated, quantified professional sanction for the attorney who relied on ChatGPT's output: $5,000 fine, direct-contempt finding, removal from the case, a bar on appearing before the court pending a disciplinary-board referral, and struck briefs; his client's appeal is delayed as a result.

How Harm Occurred

An attorney relied on a chatbot to synthesize a transcript and record into appellate briefing; the model fabricated witnesses, testimony and precedent descriptions; the unverified filing brought a judicial sanction on the user and delay to his client's appeal.

Outcome

Resolved

New Mexico Supreme Court written order, 9 September 2026: direct contempt, $5,000 fine to the State Bar Client Protection Fund within 30 days, removal from the case, bar on appearing before the court pending the disciplinary-board referral, briefs struck, appeal reassigned to the public defender (Kim Chavez Cook, appointed 2 September). The disciplinary-board process is pending.

What remains unknown

  • The 9 September 2026 order PDF was not directly inspected (nmcourts.gov search empty; form-based case lookup); the claims about it rest on two chains quoting it.
  • The underlying docket number is unconfirmed.
  • The disciplinary-board process and its outcome are pending.
  • The full set of fabricated or misdescribed citations beyond the named examples is not reported.
  • The ChatGPT product tier and whether the firm had verification policies are not reported.

What the evidence supports

AI involvement: supported. The AI link is admitted and judicially found: Aarons fed a Rev.com transcript and the record into ChatGPT (OpenAI o3), and the court's order addresses the fabricated content in the filed brief; he admitted at the show-cause hearing that he had not verified the output.

5 claims: 2 corroborated, 3 reported. What the statuses mean

Corroborated By written order of 9 September 2026, the New Mexico Supreme Court held Stephen Aarons in direct contempt, fined him $5,000 payable to the State Bar Client Protection Fund within 30 days, and referred him to the disciplinary board.

Causal attribution. The contempt, fine and referral are quoted by Ars Technica and confirmed by the Reuters wire; the ABA Journal's hearing-day account anticipated but could not confirm the later written order. The order PDF itself was not retrieved in this review.

Reported The order also removed Aarons from the case, barred him from appearing before the court pending the disciplinary-board referral, struck the filed briefs and reassigned the appeal to the public defender (Kim Chavez Cook, appointed 2 September 2026).

Causal attribution. Single-chain detail from Ars Technica's account of the order; the Reuters wire does not carry these elements.

Corroborated Aarons used ChatGPT (OpenAI o3) on a Rev.com transcript and the case record to draft the brief, which contained fabricated witnesses and testimony — officers 'Michelle Amarillo' and 'Sanchez', Manal Al-Jibury, Teresa Marquez, and false testimony attributed to Danny and Linda Stanton and Mariah Chavez — and misdescribed real precedents; he admitted at the hearing that he had not verified the output.

Causal attribution. The fabrication and the admission are established by the attorney's own statements at the hearing and the court's findings as reported; no source reports a denial.

  • arstechnica.com(opens in new tab) supports · English
    Ars passages: Rev.com transcript plus record into ChatGPT (o3); the named fabricated witnesses and false attributions; the admission of non-verification.
  • abajournal.com(opens in new tab) supports · English
    ABA Journal: 'admits stupidity over AI errors' at the show-cause hearing.
  • straitstimes.com(opens in new tab) supports · English
    Reuters headline account: 'ChatGPT invented fake police testimony in murder appeal, New Mexico high court says'.
Reported The client, Oscar Renee Sandoval, suffers a delayed appeal: the briefs were struck, Aarons was removed, and the public defender was substituted (appointed 2 September 2026); the court told Aarons 'your client is the one suffering'.

Causal attribution. Single-chain detail from Ars Technica's account of the order; the delay's length is not stated.

  • arstechnica.com(opens in new tab) supports · English
    Ars: substitution of the public defender (Kim Chavez Cook, appointed 2 Sep); the court's remark about the client.
Reported The underlying case is the murder appeal of Oscar Renee Sandoval, sentenced to life in February 2025 in Doña Ana County for killing his partner Shiereen Al-Jibury; the brief was filed around August 2025 and the State moved to strike weeks later.

Causal attribution. Background context, attributed to the wire and Ars; the docket number was not confirmed in this review.

Sources

3 sources inspected. Sources that repeat one account do not corroborate each other.

How the sources were read, and where the events happened

Read in English on 2026-09-15 (11 Sep 2026, Brodkin). Quotes the 9 September order and the 21 August hearing; names the fabricated witnesses. Applies to s1.

Read in English on 2026-09-15 (11 Sep 2026); Straits Times carrying the Reuters wire. Applies to s2.

Read in English on 2026-09-15 (24 Aug 2026). Post-hearing report citing the Santa Fe New Mexican; the fine and contempt were anticipated from the hearing, not yet the written order. Applies to s3.

Event countries: United States. Affected-person countries: United States. Court countries: United States.

Events: Santa Fe, New Mexico practice and the Doña Ana County murder case. Affected persons: the attorney and his client, both in New Mexico. Court: New Mexico Supreme Court.

Reviewed for publication 2026-09-15: Published under the 2026-09-15 charter as an own-use, work-context adverse consequence with a dated, quantified sanction and an admitted, judicially found AI link. This applies the charter's individual-review rule to the AI-sanction class; the retired ordinary_task_failure gate and a class-level ruling are not required. Aarons is a public officer of the court sanctioned by a public order and is named; his client's name appears in the public caption of the appeal. In a 2026-09-15 second pass the original compound first claim was split: the corroborated contempt/fine/referral stand as c1a, and the Ars-only order elements (removal, bar, struck briefs, reassignment) stand as reported c1b. person_relations recorded in the same pass.

People described

Stephen Aarons, Santa Fe defense attorney (the ChatGPT user, sanctioned); his client Oscar Renee Sandoval, whose murder appeal was delayed by the struck briefs

People reported harmed in this case

2 people

1 AI participant · 1 other person harmed

Two documented harmed people. The user: Aarons, sanctioned (court order reported by Ars Technica and Reuters). Another person's use consequence: his client Sandoval, whose appeal is delayed by the struck briefs and the substitution of counsel (Ars quotes the court's 'your client is the one suffering'). No further harmed persons are reported; the victim's family is not counted (no harm to them is reported).

Counted once within this case. The same person may appear in other cases. This count does not establish AI causation.

Tags

#work#justice#legal-profession#sanction#fabricated-citations#chatgpt#new-mexico

Cite this case

Compiled per our published methodology: verification statuses, sourcing standards, and corrections process.

APA

NOPE. (2026). New Mexico Supreme Court fines attorney Stephen Aarons $5,000 and removes him from a murder appeal over ChatGPT-fabricated testimony. AI incidents. https://nope.net/incidents/2026-aarons-chatgpt-fabricated-testimony-sanction

BibTeX

@misc{2026_aarons_chatgpt_fabricated_testimony_sanction,
  title = {New Mexico Supreme Court fines attorney Stephen Aarons \$5,000 and removes him from a murder appeal over ChatGPT-fabricated testimony},
  author = {NOPE},
  year = {2026},
  howpublished = {AI incidents},
  url = {https://nope.net/incidents/2026-aarons-chatgpt-fabricated-testimony-sanction}
}

Related cases

Medium ChatGPT

Brasília labour court upholds the for-cause dismissal of a manager it found used ChatGPT to alter a delivery photo to contest an iFood customer's complaint

In a judgment of 30 July 2026, the 2nd Labour Court of Brasília (Regional Labour Court of the 10th Region) upheld the dismissal for cause, on 19 March 2026, of a manager hired in June 2024. The employer said he had used artificial intelligence to falsify photos of orders on the iFood delivery app. A witness told the court she saw him use ChatGPT to alter a photo of an incomplete order, in which two of four soft drinks had been sent, so that it appeared complete, and use it to contest the customer's complaint to iFood support. The judge found this an act of dishonesty and misconduct that broke the trust required of a manager, rejected his claim that the dismissal was fabricated in retaliation for a claim he had filed the day before, dismissed all his claims, denied severance items and ordered him to pay 10% counsel fees, suspended because he was granted free legal aid. The legal news site Migalhas reported the judgment on 3 August 2026. No appeal is reported.

Medium Unidentified AI drafting tool

Massachusetts federal court revokes an out-of-state lawyer's admission and orders fees after briefs with false citations; the lawyer admitted AI drafted one

On 25 September 2026 a judge of the U.S. District Court for the District of Massachusetts granted a defence motion for sanctions under Federal Rule of Civil Procedure 11 against a California-barred lawyer admitted for one case (pro hac vice) to represent the plaintiff. The court found that an opposition the lawyer filed on 21 November 2025 was replete with inaccurate citations, including a quotation attributed to a case with a fictitious Westlaw citation, and that later oppositions cited a non-existent case. In response to the court's order to show cause, the lawyer admitted using AI to draft the first opposition and explained a belief that the enterprise-level AI software in use did not hallucinate cases. The court revoked the lawyer's pro hac vice admission and ordered the lawyer's firm to pay the defendants' reasonable fees and costs, capped at $10,000. The order does not name the AI product.

Medium First Drafts and unidentified legal research tool

Fee suit against the City of Aberdeen (N.D. Miss.): four attorneys on both sides sanctioned after AI tools produced fabricated case citations; trial cancelled and both litigants left without counsel

In a fee dispute between a Louisiana attorney and the City of Aberdeen, Mississippi, briefs filed for both sides in late 2025 cited six cases that do not exist. The attorneys admitted the citations came from unverified AI use: the plaintiff's out-of-state counsel, Kathleen M. Wilson, drafted her filing with an AI drafting program called 'First Drafts', and the City's out-of-state counsel, Kathryn Y. Williams, used an in-house AI legal research tool. Senior Judge Sharion Aycock stayed the case and cancelled the March 2026 trial, then on 8 June 2026 revoked both attorneys' pro hac vice admissions, barred them from the district for two years and fined them $2,500 and $3,500, and disqualified and fined the two local counsel who had signed the filings. Both litigants were left without counsel and given 60 days to find new representation.

High ChatGPT

Works v. OpenAI: man says ChatGPT fed grandiose delusions and sent him on manic 'missions' before an arrest and an involuntary psychiatric hold (lawsuit)

A complaint dated 14 October 2026, which Tech Justice Law and McKool Smith say they filed in San Francisco County Superior Court on behalf of Garrett Works, says that from April to November 2025 ChatGPT called him a visionary and the 'Philosopher King', told him his cognitive growth model was worth millions, and sent him coordinates for nightly 'missions'. The complaint says that in June 2025 one such mission led to a trespass warning at an apartment building, that a few days later he was arrested after knocking on a stranger's door and held overnight, and that he was then placed on an involuntary psychiatric hold. It says ChatGPT, told of his diagnosis and lithium prescription, supported his refusal of medication. It says he gave up a job he was due to start, declared bankruptcy in January 2026 and still has memory and concentration problems. The allegations are untested and OpenAI's response is not reported.

If you or someone you know is struggling, free and confidential support is available. Find a helpline near you at Signpost.

Back to all incidents